Face Generation¶
In this project, you'll use generative adversarial networks to generate new images of faces.
Get the Data¶
You'll be using two datasets in this project:
- MNIST
- CelebA
Since the celebA dataset is complex and you're doing GANs in a project for the first time, we want you to test your neural network on MNIST before CelebA. Running the GANs on MNIST will allow you to see how well your model trains sooner.
If you're using FloydHub, set data_dir to "/input" and use the FloydHub data ID "R5KrjnANiKVhLWAkpXhNBe".
data_dir = '/data'
!pip install matplotlib==2.0.2
# FloydHub - Use with data ID "R5KrjnANiKVhLWAkpXhNBe"
#data_dir = '/input'
"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
import helper
helper.download_extract('mnist', data_dir)
helper.download_extract('celeba', data_dir)
show_n_images = 25
"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
%matplotlib inline
import os
from glob import glob
from matplotlib import pyplot
mnist_images = helper.get_batch(glob(os.path.join(data_dir, 'mnist/*.jpg'))[:show_n_images], 28, 28, 'L')
pyplot.imshow(helper.images_square_grid(mnist_images, 'L'), cmap='gray')
CelebA¶
The CelebFaces Attributes Dataset (CelebA) dataset contains over 200,000 celebrity images with annotations. Since you're going to be generating faces, you won't need the annotations. You can view the first number of examples by changing show_n_images.
show_n_images = 25
"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
mnist_images = helper.get_batch(glob(os.path.join(data_dir, 'img_align_celeba/*.jpg'))[:show_n_images], 28, 28, 'RGB')
pyplot.imshow(helper.images_square_grid(mnist_images, 'RGB'))
Preprocess the Data¶
Since the project's main focus is on building the GANs, we'll preprocess the data for you. The values of the MNIST and CelebA dataset will be in the range of -0.5 to 0.5 of 28x28 dimensional images. The CelebA images will be cropped to remove parts of the image that don't include a face, then resized down to 28x28.
The MNIST images are black and white images with a single color channel while the CelebA images have 3 color channels (RGB color channel).
Build the Neural Network¶
You'll build the components necessary to build a GANs by implementing the following functions below:
model_inputsdiscriminatorgeneratormodel_lossmodel_opttrain
Check the Version of TensorFlow and Access to GPU¶
This will check to make sure you have the correct version of TensorFlow and access to a GPU
"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
from distutils.version import LooseVersion
import warnings
import tensorflow as tf
# Check TensorFlow Version
assert LooseVersion(tf.__version__) >= LooseVersion('1.0'), 'Please use TensorFlow version 1.0 or newer. You are using {}'.format(tf.__version__)
print('TensorFlow Version: {}'.format(tf.__version__))
# Check for a GPU
if not tf.test.gpu_device_name():
warnings.warn('No GPU found. Please use a GPU to train your neural network.')
else:
print('Default GPU Device: {}'.format(tf.test.gpu_device_name()))
Input¶
Implement the model_inputs function to create TF Placeholders for the Neural Network. It should create the following placeholders:
- Real input images placeholder with rank 4 using
image_width,image_height, andimage_channels. - Z input placeholder with rank 2 using
z_dim. - Learning rate placeholder with rank 0.
Return the placeholders in the following the tuple (tensor of real input images, tensor of z data)
import problem_unittests as tests
def model_inputs(image_width, image_height, image_channels, z_dim):
"""
Create the model inputs
:param image_width: The input image width
:param image_height: The input image height
:param image_channels: The number of image channels
:param z_dim: The dimension of Z
:return: Tuple of (tensor of real input images, tensor of z data, learning rate)
"""
# TODO: Implement Function
real_images = tf.placeholder(tf.float32,shape=(None,image_height,image_width,image_channels),name='real_images')
z_input = tf.placeholder(tf.float32,shape=(None,z_dim),name='z_input')
learning_rate = tf.placeholder(tf.float32,name='learning_rate')
return real_images, z_input, learning_rate
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_model_inputs(model_inputs)
Discriminator¶
Implement discriminator to create a discriminator neural network that discriminates on images. This function should be able to reuse the variables in the neural network. Use tf.variable_scope with a scope name of "discriminator" to allow the variables to be reused. The function should return a tuple of (tensor output of the discriminator, tensor logits of the discriminator).
Discriminator Solution¶
For my first pass at implementation, I'm following the advice in the DCGAN paper:
- No max pooling between convolutions; instead use strided convolutions
- No fully connected layer between final convolution and output
- Batch norm on all except the first convolution layer
- Leaky ReLU on all except output layer
- Sigmoid on output layer
def discriminator(images, reuse=False, alpha=0.1, is_training=False):
"""
Create the discriminator network
:param images: Tensor of input image(s)
:param reuse: Boolean if the weights should be reused
:return: Tuple of (tensor output of the discriminator, tensor logits of the discriminator)
"""
# TODO: Implement Function
with tf.variable_scope('discriminator',reuse=reuse):
# initializer = tf.contrib.layers.xavier_initializer()
# Input layer is 28x28x3 (celebs) or 28x28x1 (MNIST)
# convolutional layer for input with leaky ReLU (no batch norm)
x1 = tf.layers.conv2d(images,64,5,strides=(2,2),padding='same', kernel_initializer=tf.contrib.layers.xavier_initializer())
x1 = tf.maximum(x1, x1 * alpha)
x1 = tf.layers.dropout(x1,0.1,training=is_training)
# x2 = tf.layers.conv2d(x1,64,1,strides=(2,2),padding='same', kernel_initializer=initializer)
# x2 = tf.layers.batch_normalization(x2,training=is_training)
# x2 = tf.maximum(x2, x2 * alpha)
# x2 = tf.layers.dropout(x2,0.1,training=is_training)
#now 14x14x64
# strided convolutional layer with batch norm and leaky ReLU
x3 = tf.layers.conv2d(x1,128,5,strides=(2,2),padding='same', kernel_initializer=tf.contrib.layers.xavier_initializer())
x3 = tf.layers.batch_normalization(x3,training=is_training)
x3 = tf.maximum(x3, x3 * alpha)
x3 = tf.layers.dropout(x3,0.1,training=is_training)
# x4 = tf.layers.conv2d(x3,128,1,strides=(2,2),padding='same', kernel_initializer=initializer)
# x4 = tf.layers.batch_normalization(x4,training=is_training)
# x4 = tf.maximum(x4, x4 * alpha)
# x4 = tf.layers.dropout(x4,0.1,training=is_training)
# now 7 x 7 x 128
# strided convolutional layer with batch norm and leaky ReLU
x5 = tf.layers.conv2d(x3,256,5,strides=(2,2),padding='same', kernel_initializer=tf.contrib.layers.xavier_initializer())
x5 = tf.layers.batch_normalization(x5,training=is_training)
x5 = tf.maximum(x5, x5 * alpha)
# x5 = tf.layers.dropout(x5,0.1,training=is_training)
# x6 = tf.layers.conv2d(x5,256,1,strides=(2,2),padding='same', kernel_initializer=initializer)
# x6 = tf.layers.batch_normalization(x6,training=is_training)
# x6 = tf.maximum(x6, x6 * alpha)
# x6 = tf.layers.dropout(x6,0.1,training=is_training)
# now 4 x 4 x 256 (since padding='same')
flat = tf.reshape(x5, (-1, 4*4*256))
logits = tf.layers.dense(flat,1)
out = tf.sigmoid(logits)
return out, logits
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_discriminator(discriminator, tf)
Generator¶
Implement generator to generate an image using z. This function should be able to reuse the variables in the neural network. Use tf.variable_scope with a scope name of "generator" to allow the variables to be reused. The function should return the generated 28 x 28 x out_channel_dim images.
Generator Solution¶
Again, following the DCGAN paper:
- layers of fractionally-strided convolutions
- no fully connected layers
def generator(z, out_channel_dim, is_train=True, alpha=0.2):
"""
Create the generator network
:param z: Input z
:param out_channel_dim: The number of channels in the output image
:param is_train: Boolean if generator is being used for training
:return: The tensor output of the generator
"""
# TODO: Implement Function
reuse = not(is_train)
with tf.variable_scope('generator', reuse=reuse):
# initializer = tf.contrib.layers.xavier_initializer()
# Input layer is 28x28x3 (celeba) or 28x28x1 (MNIST)
# convolutional layer for input with leaky ReLU (no batch norm)
x1 = tf.layers.dense(z,4*4*1024) # fully connected layer to start
x1 = tf.reshape(x1,(-1,4,4,512)) # reshape to match the size for convolutional net
x1 = tf.maximum(alpha * x1, x1 ) # leaky ReLU
# x1 = tf.layers.dropout(x1,0.1,training=is_train)
x2 = tf.layers.conv2d_transpose(x1,1024,1,strides=1,padding='same',kernel_initializer=tf.contrib.layers.xavier_initializer())
x2 = tf.layers.batch_normalization(x2,training=is_train) # batch normalization
x2 = tf.maximum(alpha * x2, x2 ) # leaky ReLU
# Now 4x4x1024
x3 = tf.layers.conv2d_transpose(x2,512,4,strides=1,padding='valid',kernel_initializer=tf.contrib.layers.xavier_initializer())
x3 = tf.layers.batch_normalization(x3,training=is_train) # batch normalization
x3 = tf.maximum(alpha * x3, x3 ) # leaky ReLU
# Now 7x7x512
x4 = tf.layers.conv2d_transpose(x3,512,1,strides=1,padding='same',kernel_initializer=tf.contrib.layers.xavier_initializer())
x4 = tf.layers.batch_normalization(x4,training=is_train) # batch normalization
x4 = tf.maximum(alpha * x4, x4 ) # leaky ReLU
x5 = tf.layers.conv2d_transpose(x4,256,5,strides=2,padding='same',kernel_initializer=tf.contrib.layers.xavier_initializer())
x5 = tf.layers.batch_normalization(x5,training=is_train) # batch normalization
x5 = tf.maximum(alpha * x5, x5 ) # leaky ReLU
# Now 14x14x256
x6 = tf.layers.conv2d_transpose(x5,256,1,strides=1,padding='same',kernel_initializer=tf.contrib.layers.xavier_initializer())
x6 = tf.layers.batch_normalization(x6,training=is_train) # batch normalization
x6 = tf.maximum(alpha * x6, x6 ) # leaky ReLU
x7 = tf.layers.conv2d_transpose(x6,128,5,strides=2,padding='same',kernel_initializer=tf.contrib.layers.xavier_initializer())
x7 = tf.layers.batch_normalization(x7,training=is_train) # batch normalization
x7 = tf.maximum(alpha * x7, x7 ) # leaky ReLU
# Now 28x28x128
logits = tf.layers.conv2d_transpose(x7,out_channel_dim,5,strides=1,padding='same')
# Now 28x28x(out_channel_dim)
# Squash output to (-1,1) with tanh
out = tf.tanh(logits)
return out
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_generator(generator, tf)
Loss¶
Implement model_loss to build the GANs for training and calculate the loss. The function should return a tuple of (discriminator loss, generator loss). Use the following functions you implemented:
discriminator(images, reuse=False)generator(z, out_channel_dim, is_train=True)
def model_loss(input_real, input_z, out_channel_dim, alpha=0.2, smooth=0.1):
"""
Get the loss for the discriminator and generator
:param input_real: Images from the real dataset
:param input_z: Z input
:param out_channel_dim: The number of channels in the output image
:return: A tuple of (discriminator loss, generator loss)
"""
# TODO: Implement Function
g_model = generator(input_z,out_channel_dim,alpha=alpha)
d_model_real, d_logits_real = discriminator(input_real,reuse=False,alpha=alpha)
d_model_fake, d_logits_fake = discriminator(g_model,reuse=True,alpha=alpha)
d_loss_real = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(
logits=d_logits_real,
labels=tf.ones_like(d_model_real)*(1-smooth)
))
d_loss_fake = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(
logits=d_logits_fake,
labels=tf.zeros_like(d_model_fake)*(1-smooth) # note smoothing doesn't do anything here
))
g_loss = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(
logits=d_logits_fake,
labels=tf.ones_like(d_model_fake)
))
d_loss = d_loss_real + d_loss_fake
return d_loss, g_loss
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_model_loss(model_loss)
Optimization¶
Implement model_opt to create the optimization operations for the GANs. Use tf.trainable_variables to get all the trainable variables. Filter the variables with names that are in the discriminator and generator scope names. The function should return a tuple of (discriminator training operation, generator training operation).
def model_opt(d_loss, g_loss, learning_rate, beta1):
"""
Get optimization operations
:param d_loss: Discriminator loss Tensor
:param g_loss: Generator loss Tensor
:param learning_rate: Learning Rate Placeholder
:param beta1: The exponential decay rate for the 1st moment in the optimizer
:return: A tuple of (discriminator training operation, generator training operation)
"""
# Get variables
t_vars = tf.trainable_variables()
g_vars = [var for var in t_vars if var.name.startswith('generator')]
d_vars = [var for var in t_vars if var.name.startswith('discriminator')]
# Optimization
with tf.control_dependencies(tf.get_collection(tf.GraphKeys.UPDATE_OPS)):
g_train_opt = tf.train.AdamOptimizer(learning_rate,beta1).minimize(g_loss,var_list=g_vars)
d_train_opt = tf.train.AdamOptimizer(learning_rate,beta1).minimize(d_loss,var_list=d_vars)
return g_train_opt, d_train_opt
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_model_opt(model_opt, tf)
"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
import numpy as np
def show_generator_output(sess, n_images, input_z, out_channel_dim, image_mode):
"""
Show example output for the generator
:param sess: TensorFlow session
:param n_images: Number of Images to display
:param input_z: Input Z Tensor
:param out_channel_dim: The number of channels in the output image
:param image_mode: The mode to use for images ("RGB" or "L")
"""
cmap = None if image_mode == 'RGB' else 'gray'
z_dim = input_z.get_shape().as_list()[-1]
example_z = np.random.normal(0.0, 1.0, size=[n_images, z_dim])
samples = sess.run(
generator(input_z, out_channel_dim, False),
feed_dict={input_z: example_z})
images_grid = helper.images_square_grid(samples, image_mode)
pyplot.imshow(images_grid, cmap=cmap)
pyplot.show()
Train¶
Implement train to build and train the GANs. Use the following functions you implemented:
model_inputs(image_width, image_height, image_channels, z_dim)model_loss(input_real, input_z, out_channel_dim)model_opt(d_loss, g_loss, learning_rate, beta1)
Use the show_generator_output to show generator output while you train. Running show_generator_output for every batch will drastically increase training time and increase the size of the notebook. It's recommended to print the generator output every 100 batches.
import matplotlib.pyplot as plt
%matplotlib inline
def train(epoch_count, batch_size, z_dim, learning_rate, beta1, get_batches, data_shape, data_image_mode,alpha=0.2):
"""
Train the GAN
:param epoch_count: Number of epochs
:param batch_size: Batch Size
:param z_dim: Z dimension
:param learning_rate: Learning Rate
:param beta1: The exponential decay rate for the 1st moment in the optimizer
:param get_batches: Function to get batches
:param data_shape: Shape of the data
:param data_image_mode: The image mode to use for images ("RGB" or "L")
"""
# TODO: Build Model
image_height = data_shape[1]
image_width = data_shape[2]
image_channels = len(data_image_mode) # RGB = 3; L = 1
real_images, z_input, learn_rate = model_inputs(image_width, image_height, image_channels, z_dim)
d_loss, g_loss = model_loss(real_images, z_input, image_channels, alpha=alpha)
g_train_opt, d_train_opt = model_opt(d_loss, g_loss, learning_rate, beta1)
steps = 0
print_every = 10
show_every = 100
d_loss_vec = []
g_loss_vec = []
saver = tf.train.Saver()
batches_per_epoch = data_shape[0]//batch_size
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
for epoch_i in range(epoch_count):
steps = 0
for i,batch_images in enumerate(get_batches(batch_size)):
steps +=1
batch_images *= 2.
batch_z = np.random.normal(0.0,1.0,size=(batch_size,z_dim))
d, _ = sess.run([d_loss, d_train_opt],feed_dict={real_images:batch_images, z_input:batch_z, learn_rate:learning_rate})
g, _ = sess.run([g_loss, g_train_opt],feed_dict={z_input:batch_z, learn_rate:learning_rate})
d_loss_vec.append(d)
g_loss_vec.append(g)
if steps % print_every == 0:
train_loss_d = d_loss.eval({real_images:batch_images, z_input:batch_z, learn_rate:learning_rate})
train_loss_g = g_loss.eval({z_input:batch_z, learn_rate:learning_rate})
print("Epoch {}/{}...".format(epoch_i+1,epoch_count),
"Step {}/{}...".format(i+1,batches_per_epoch),
"Discriminator Loss {:.4f}".format(train_loss_d),
"Generator Loss {:.4f}".format(train_loss_g)
)
if steps % show_every == 0:
show_generator_output(sess,16,z_input,image_channels,data_image_mode)
saver.save(sess, './checkpoints/generator.ckpt')
show_generator_output(sess,8,z_input,image_channels,data_image_mode)
Discriminator_loss, = plt.plot(d_loss_vec, color='b', label='Discriminator loss')
Genereator_loss, = plt.plot(g_loss_vec, color='r', label='Generator loss')
plt.legend(handles=[ Discriminator_loss, Genereator_loss])
MNIST¶
Test your GANs architecture on MNIST. After 2 epochs, the GANs should be able to generate images that look like handwritten digits. Make sure the loss of the generator is lower than the loss of the discriminator or close to 0.
batch_size = 128
z_dim = 200
learning_rate = 0.0008
beta1 = 0.2
alpha = 0.1
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
epochs = 2
mnist_dataset = helper.Dataset('mnist', glob(os.path.join(data_dir, 'mnist/*.jpg')))
with tf.Graph().as_default():
train(epochs, batch_size, z_dim, learning_rate, beta1, mnist_dataset.get_batches,
mnist_dataset.shape, mnist_dataset.image_mode)
CelebA¶
Run your GANs on CelebA. It will take around 20 minutes on the average GPU to run one epoch. You can run the whole epoch or stop when it starts to generate realistic faces.
batch_size = 32
z_dim = 200
learning_rate = 0.0004
beta1 = 0.2
alpha = 0.1
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
epochs = 1
celeba_dataset = helper.Dataset('celeba', glob(os.path.join(data_dir, 'img_align_celeba/*.jpg')))
with tf.Graph().as_default():
train(epochs, batch_size, z_dim, learning_rate, beta1, celeba_dataset.get_batches,
celeba_dataset.shape, celeba_dataset.image_mode)
Submitting This Project¶
When submitting this project, make sure to run all the cells before saving the notebook. Save the notebook file as "dlnd_face_generation.ipynb" and save it as a HTML file under "File" -> "Download as". Include the "helper.py" and "problem_unittests.py" files in your submission.